Smart Checkout System

Abstract

This study introduces an intelligent Smart Checkout System based on state-of-the-art computer vision and deep learning technology to streamlines the retail transaction process. The system eliminates the traditional barcode scanning with the use of real-time instance segmentation for facilitate identification of products and billing. Adopting the state-of-the-art YOLO architectures (YOLOv8 and YOLOv11), the system can achieve high-precision detection under multiple scenarios of retail conditions including occlusions, varying lighting conditions, and complex product grouping. One private data set of 1,525 well-annotated product images was collected and utilized to train and test the models to impart real-world robustness. Experimental findings verify improved performance, where YOLOv11 achieved a mean Average Precision (mAP50) value of 0.97 and YOLOv8 achieved 0.95, while simultaneously retaining computational efficiency. The system addresses major retail automation challenges, such as inventory management, shoplifting prevention, and cost of operations reduction, with a focus on flexibility to accommodate regional market needs (e.g., cash economies in the Arab region). The primary contributions

are a scalable deployment pipeline, model trade-off comparison, and deployment strategies for high-accuracy and resource- constrained environments. The project also aligns theoretical innovation with practical purpose, offering a model for retail

modernization in emerging markets through AI.